The dirty little secret about silicon’s successors

Graphene-like 2D materials will never live up to their promise in electronics if researchers don’t start focusing on growing high-quality layers, Cambridge researchers argue.

Paul van Gerven

Transition-metal dichalcogenides (TMDs) have become serious contenders to replace silicon in future chips. Several properties have prompted leading firms such as Imec and TSMC to look into TMD monolayers as a way to keep transistor innovation going once silicon’s decades-long run hits a wall.

For the 2D materials to find their way into the fab, however, challenges remain. In an invited paper in Nature Electronics, Yan Wang and colleagues from the University of Cambridge argue that most of these can be traced back to a single issue: the poor quality of TMD monolayers. It’s a hurdle “the community has been reluctant to discuss openly in the context of performance of electronic devices,” the authors write.

Artist impression of transistors made from 2D materials. Credit: Shanxi University

Low leakage

The Nobel Prize-winning discovery of graphene almost twenty years ago set off a flurry of activity to exploit the new material’s properties. In microelectronics, graphene has two things going for it. First, it’s as thin as it gets: one atom thick. Since the properties of bulk semiconductors start to break down as you make thinner layers, silicon could never match that. Secondly, electrons zigzag easily through graphene’s crystal lattice, making for faster electronics.

For all its promise, however, graphene never made waves in electronics. That’s largely due to it being a semimetal, not a semiconductor. It’s possible to induce the bandgap needed for electronic switching action in graphene, but once natively semiconducting TMDs emerged, they seemed a more practical alternative. TMDs aren’t quite as thin as graphene, but three atomic layers – essentially, elements from the oxygen family sandwiching a transition metal – is still well beyond what silicon could muster.

TMDs feature several properties that chipmakers appreciate. A low subthreshold swing makes for electronic switches that turn on and off with a relatively small change in gate voltage. This helps reduce power consumption. A high on/off ratio improves power efficiency as well, thanks to low leakage current in the off state. The electron mobility in TMDs, on the other hand, tends to be much lower than in graphene and not necessarily better than in silicon. Fortunately, it’s still sufficient – and there are ways to improve it. Which brings us to the point of the paper in Nature Electronics.

Mandatory

Leveraging the favorable properties of TMDs to build great next-gen transistors hinges on the availability of high-quality 2D TMD monolayers, the Cambridge researchers point out. Monolayers grown by methods compatible with industrial manufacturing are “generally known” to be “riddled” with defects, which “detrimentally affect device properties, including on-state current, variation in threshold voltage and mobility.” Defects also complicate doping and developing crucial other ingredients of TMD transistors, such as suitable gate dielectrics and contacts.

Researchers routinely claim to have grown high-quality material but offer little evidence to back up that claim. The Cambridge paper argues for a change in attitude of sorts, encouraging to focus more on improving quality and less on reporting yet another slight variation of existing work. And inclusion of data on defect density should be mandatory, of course.